Index Home About Blog
From: (Floyd Davidson)
Newsgroups: comp.dcom.modems,
Subject: Re: Slower modem connections thru T1/channel bank
Date: 22 Feb 1996 07:23:16 GMT wrote:
>The ISP I use has also been bitten by the dreaded "channel bank" problem. 
>After rigorous testing with BellSouth (the local RBOC) and US Robotics (the
>ISP uses USR Total Control racks), it was finally determined that the channel
>banks are somehow inducing some digital "noise" which thereby limits the
>connect speed.  As a test, the RBOC put in one T1 line direct off of the
>switch to a TC rack and connects on that are running excellent, with a lot of
>people getting consistant 28.8, 31.2 and 33.6 connection rates.  Of course,
>the straight T1 is a LOT more expensive than the ones coming from the channel
>banks, which are a curse to society, IMHO.

Your observations and conclusions are right on the target.

  +------+                     +-----------+
  | modem|_____analog____//____|     CO    |_______\    digital to 
  | bank |      line    //     | line card |       /    the pstn
  +------+                     +-----------+

          \___ significant ___/
               analog loop

A normal telephone line...  The length of the analog loop more or
less determines the bandwidth of the entire circuit.  The AD-DA
conversion in the "CO line card" determines the absolute best
possible S/N because it generates error noise (quanitization
noise) equal to a S/N of 37 dB.  It is likely that the
analog loop S/N is not that good to start with...

The problem is that noise adds up, but the signal level does not.
So if the analog loop has a given level of noise and the CO line
card has exactly the same level of noise, the total is 3dB higher
than either alone.  (If the two are different by a greater amount,
the lesser one adds less, and anything more than 6 dB different
can be ignored for practical purposes.)

In the specific case where the two noise sources are equally at a
37 dB S/N, the combined noise will give a S/N of 34 dB.  I don't
have v.34 specs at hand, but if memory servers that will allow a
28.8Kbps connection with a couple dB to spare.

Now lets add this extra channel bank in:

At the customer end:

  +------+                     +-----------+
  | modem|_____analog____//____| Chan Bank |_______\    digital to 
  | bank |      line    //     | FXS card  |       /    central office
  +------+                     +-----------+

          \___ very _short_ ___/
               (very _good_)
               analog loop

At the CO end:

     +-----------+        +-----------+
  \__| Chan Bank |________|     CO    |_______\    digital to 
  /  | FXO card  |        | line card |       /    the pstn
     +-----------+        +-----------+

digital            analog
facility            link

In this case we have a channel bank (with an additional AD-DA
conversion which adds quanitization noise).  The extra noise
sets the absolute _best_ S/N at 34 dB (if the entire rest of
the circuit is significantly better than that, which is not
likely).  Most likely the rest of the circuit will all total
add just about that same amount of noise, and the result is
maybe about a 31 dB S/N and instead of a 28.8 Kbps connection
the modems connect at 26.4Kbps, at best.

If the digital facility from the customer location goes directly
to a digital interface to the switch, thus skipping the extra
AD-DA conversion, the quanitization noise is reduced by as much as
3 dB on really good lines, and faster connections occur.

Now, with all that said...  understand that the telco
specification for S/N is 24 dB or better for normal lines.  So
called data conditioned lines sometimes can be arranged for, and
they still only guarantee a 28 dB S/N.  All of which is to say
that in either configuration it should be obvious that the
customer is getting far better parameters than is required.
Indeed, a worst case digital facility using channel banks might be
significantly better than an all copper/analog loop if that loop
is just an average loop!

I'm not at all sure why the telco would charge more for a digital
interface at the CO, or for that matter why they would install the
extra equipment in the first place and force a customer to use the
channel bank analog interface!


Floyd L. Davidson          Salcha, Alaska

From: (Floyd Davidson)
Newsgroups: comp.dcom.modems
Subject: Re: Phone Company Poor Performance
Date: 27 Aug 1998 05:02:10 GMT

In article <6s1hss$1cd6$>, Hooda Gest <> wrote:
>Steve Uhrig wrote in message <6rvg7s$3p6$>...

>>    The extra D/A conversion is there because the telco has to
>>purchase equipment that can provide the most
>>versatillity. Direct interfaced carriers can't be used for
>>special service applications such as 56-64K digital point to
>>point service, FX service, fractional T1 service etc.
>>Carriers with COTs can be used for practically any service by
>>equippingit with the proper channel plugs. Interfaced carrier
>>is only good for POTS service. Another point is that most COs
>>only have a very few carriertypes that will interface with the
>>switch. The telco will buy whatevercarrier they can get for
>>the lowest price that will give them the most
>>flexibillity. This is not usually the type that interfaces
>>with the switch.
>Universal interfacing means that there is extra equipment between the
>digital line of the customer and the central office switch. To be

>>>It is the expense of converting existing systems which the
>>>telcos appear hesitant to underwrite.
>>They don't want to convert because the lose flexabillity by
>>interfacing COTed systems.

>Flexibility is important. But what kind and for what purpose? I am on an
>integrated SLC. I have had cable pair changes due to trouble and they
>are easy since they only need to be done at the field end, not at both
>field and CO. Changes at the CO can be done electronically since it is
>all dedicated. If the SLC is integrated then there is no equipment at
>the CO to change, no cable pair to change, no nada... The customer is
>just another trunk (channel) in the system. Failures are possible for a
>single customer only in the field at the SLC head and on the customer's
>line from there into the home.

Steve has a very good point, which appears to have been missed.
The flexibility he is speaking of is in the functionality of the
installed facilities.  An SLC is essentially a trunking system
from a remote location to the CO, and if an integrated interface
is used then each DS1 interface (each T1, with 24 channels) is
_dedicated_ to being used for whatever single service is
available on the CO DS1 interface that it is connected to.
Typically the discussion is centered on POTS service, and the
entire group of 24 channels would be available for no other
service.  Telco's also have leased line services in many areas,
and that is what Steve as refering to, plus special types of
POTS service and ISDN too.  If the SLC terminates directly into
a CO switch...  then another T1 line has to be used to provide
other services, and that is expensive.  By terminating the T1
into a channel bank the telco can then use any channel for any
service.  (If we are talking about a remote location that has
only 1 to 5 T1's, that is significant.  If it has 20 then
dedicating each of them is just not a problem.)

However, universal interfacing for SLC's is still poor
engineering, and generally happens because managers and
engineers tend to be trapped in old paradigms.  The _right_ way
to provide flexibility for facility utilization (and cut
maintenance costs, increase reliability, and centralize work
load and functionality) is to install a digital cross connect
system and interface _all_ DS1's into it (SLC's and CO switching
systems, as well as interswitch trunks and any other T1
facilities).  They can then reduce the on site work force and
maintain facility assignments with _clerks_ from a remote
location rather than having expensive technical staff doing it
on location.

Since I _am_ technical staff, I've not always been exceedingly
happy to see it happen, but I have to admit that it is the
appropriate design philosophy.


Floyd L. Davidson                      
Ukpeagvik (Barrow, Alaska)

Index Home About Blog